ACEBench
text
+
+
+
+
About
ACEBench is a comprehensive AI benchmark for evaluating Large Language Models' tool usage capabilities across three scenarios: Normal (basic tool usage), Special (ambiguous instructions), and Agent (multi-agent interactions). Covering 8 major domains and 68 sub-domains, it tests LLMs' decision-making and reasoning when integrated with various tools through 1-8 dialogue turns simulating real-world contexts. The benchmark addresses limitations in existing evaluations by providing granular error analysis.
+
+
+
+
Evaluation Stats
Total Models2
Organizations1
Verified Results0
Self-Reported2
+
+
+
+
Benchmark Details
Max Score1
Language
en
+
+
+
+
Performance Overview
Score distribution and top performers
Score Distribution
2 models
Top Score
76.5%
Average Score
76.5%
High Performers (80%+)
0Top Organizations
#1Moonshot AI
2 models
76.5%
+
+
+
+
Leaderboard
2 models ranked by performance on ACEBench
| License | Links | ||||
|---|---|---|---|---|---|
| Jul 11, 2025 | MIT | 76.5% | |||
| Sep 5, 2025 | MIT | 76.5% |
